http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=120734745699
Does anyone with EE skills thinks those can be used to make some kind of (bigger) 3A driver? I know it's not current regulator, but still....?
I took a look at the circuit and it looks like it could be done fairly simply. But "fairly simply" has its cost...
It will require two to five cells (4.75v to 23V), and may not support PWM dimming to well.
The best efficiency (estimate) will be about 78% using two cells, efficiency will decrease slightly with more cells.
Reference the basic schematic (on the eBay page or the monolithicpower.com website).
Connect the + side of a XM-L emitter to the + output, connect the - side to a 0.1ohm 1% 1Watt resistor. Connect the other side of this resistor to Ground. Remove the resistor R2. Connect one side a 6.04k ohm 1% resistor to where R2 connected to R1 and pin5 of the MP2307. Connect the other side to the - side of the XM-L (where it connects to the 0.1 ohm).
OK, have I lost you yet? I hope not, because we're done!
Low frequency PWM (less than 200-300Hz) could work by driving the Enable pin7, (R4 may need to be removed).
It's simple, a bit messy, and a bit large, but it could just work. Or, you may want to keep looking for something dedicated to driving LEDs.
When you try something new, you learn something. (sometimes you learn that you shouldn't do that)
I'd try to avoid making a voltage regulator into a current regulator... but it's definitely doable.
Since the feedback voltage is .925v, the only simple way to accurately use it as a 3A constant-current regulator (for driving XM-Ls, P7s, etc) would be a with a .33 ohm resistor for the current feedback, dissipating 3 watts, needing a small wire-wound or a very large carbon film.
Crux: I'm not sure how well that idea would work... Seems like you'd be building a ~5V regulator (.925/((6040+.1)/(6040+.1+26100))) with a very, very slight amount of current sensing? Also note the board is obviously not that schematic - that's just the reference example copied from the PDF.
--Bushytails
Thanks Crux! Well, I'm lost. If you could do the sketch and publish it somehow I might understand ;-)
Anyway, my intension is to use 8.4V or 13.2V (at their max V) battery packs and more than one XM-L. I would prefer solution with 3 XM-L because I could use somewhat wide lenses. I already have 8.4 battery pack (+ DIY spare one) and charger from MJ-808 bicycle lamp, so I would prefer to use those, but yes I know - it might be to low voltage, isn't it?
Do you know any relatively cheap solution to drive those leds with 8.4V pack (2S2P 18650), up to 2.4A (max. 3A), multimode (maybe PWM) or pot regulated to get different brightnes? It doesn't to be very small as I can design my own housing up to 35mm diameter.
To drive 3x XM-L in series, you need >9V output, so for a 8.4V pack, you'd need a boost converter, not a buck converter like above. In parallel, you'd need a 9A output, not a 3A output. It might work with a 3-cell pack, but it'd be close, especially as the cells near empty.
The easiest option is probably three separate flashlight buck drivers - they're not too expensive from the usual chinese dealers, although they're a lot more than the converter above.
If you use a 4-cell pack with them all in series, the above converter would work well, although the efficiency still wouldn't be that great.
--Bushytails
Ledoman,
Yes, it would be nice to make a driver that could do all those things - it could be done - but it would take some investment and engineering time to accomplish. Two commodities I don't have in abundance. I think Bushytails' idea of using three drivers, one for each LED, is a good one. That way you could select one, two or three LEDs at once. I'd parallel the batteries to increase runtime - 2P1S. Buying several ready made drivers is a sure bet. Trying to make something work that you don't quite understand (yet) may lead to frustration, wasted time and money. Or... maybe you'll find success. I don't know your skill, resources or determination. I've been looking at using a circuit a lot like this one for a torch I'm making, yet I keep going back to see if there is a pre-made driver that would save me the trouble.
Keep in mind that 3 LEDs @ 3A = 30Watts (roughly) - that's a lot of power for 18650 cells. Assuming that you want this to last an hour, you would need 5 x 2Ah cells. (5 x 3.6V(typ. voltage) x 2Ah) = 36Wh Figure 80% efficiency and your left with ~29W to the LEDs. Now don't forget about the heat generated by all this power... fins or fans, or both.
You could also use a single SST-90 driver and parallel them.
--Bushytails
Bushytails,
If I get some time tonight I'll post a diagram of what I am talking about. I've used this scheme in several lights I've designed.
One explanation can be found here:
http://focus.ti.com/general/docs/litabsmultiplefilelist.tsp?literatureNumber=slva364
See page 2. They're applying it to a TPS61200 IC but the concept will work with most anything else...
That would be dependent on the output voltage, not an actual constant current supply... It'd mostly be a voltage regulator. You'd need to stick a 7805 or similar in there for a regulated voltage, and it's no longer simple. :)
--Bushytails
At $9 it is hard to argue with the SST-90 driver. Give it three cells and it gives you 6.5Amps in 5 modes. Done!
Thanks guys. I know 3 cells would be the best, but as I said I already have 2S2P cells pack and charger for it. In case 3S2P I would have to invest also into those two and charger needs balancing feature. That can easily cost up to 50$ or more (guesstimating).
I've found some interesting stuff here http://www.candlepowerforums.com/vb/showthread.php?201392-Poorman-Mutli-Lux-setup-method - still 1 led - 1 cell relation. Based on that I've got an idea which is not most efficient, but could be poorman solution 
.
So we got 2 cells, 3 leds and one 8*AMC7513 driver. I could connect first two leds in paralell and third serialy (over driver). That should drive first two leds at 1.4A and third at 2.8A. This will also spare some energy and lower the heat. Shure, at 1.4A XM-Ls will be underdriven, but illumination should still be quite good.
What is to concern is lower Vf of the XM-Ls (comparing to original design for P7) and extra energy at higher voltages which would transform to heat at the driver. Have seen those concerns somewhere, but can't recall where.
Wondering also how Sky Ray 3*T6 is driven as there is same problem except if they have boost driver.
.
Wondering also how Sky Ray 3*T6 is driven as there is same problem except if they have boost driver.
That one looks like it's using boost driver, and the 3*T6 "SOZTMxxxx" one is using buck driver. There forum members w/ both so maybe ask in their respective threads.
In terms of mechanical design, I've been thinking of putting them on a round computer heatsink and sticking that inside some pvc. Batteries should provide ~ right voltage for the fan and I just stuff the electronics and batteries upstream of it. What's been holding me up is finding just the right size one to go in a 3" pipe.
Crux, I take it you're talking about this,
http://www.dealextreme.com/p/12v-sst-90-5-mode-memory-circuit-board-for-flashlights-50026?r=43033281
not this?
I've seen this wiring method before. And while it can work, there two things to look out for.
- You need to be sure to heat-sink the driver. The 7135's work by dropping the excess voltage difference between the batteries and the LEDs. When the batteries are at 4.2V and the LED drops 3.3V at 2.8A there is a difference of 0.9V across the 7135's output pin. And the driver will burn 0.9V x 2.8A = 2.5 Watts. This is the normal setup.
When there are 2 cells and 2 LEDs (but still 1 driver) then the voltage difference and the power dissipation is double - 5Watts. Think of the heat-sink you would give to a 5W led and that is what the driver alone needs. Thankfully the cell voltage doesn't stay at 4.2V for long so the dissipation drops off too. - Over-voltage of the 7135's and/or uController. When running 2 cells and 2 LEDs the voltage at the driver power input is 8.4V - 3.3V = 5.1V. This is fine. But if using 3 cells and 3 LEDs then the voltage is 12.6V - 6.6V = 6V, the upper limit for the 7135's and the uController. In this case making sure the diode on the driver board is a standard silicon type (Vf = 0.6V) not a low voltage Schottky (Vf = 0.2V) will help. You should be fine with 2 cells.
If you choose this design then I would wire the single LED in series with the driver which is driving the other two in parallel. All this gets you is one less point of failure, if any of the series LEDs should short, the driver will fry.
Good luck!
Agenthex,
Yes, the DX one. I didn't see the one at KD, and it's claim is 9A... nice...
I'm trying to visualize your setup... can't quite do it...
I'm still looking for a reason to use this: :O
I have that exact one, lol. But I have a whole bunch HP made for its mainframes/servers that I bought on the cheap back in the day but never used. Unfortunately they all seem a tad bit bigger than the 2"/3" commonly available pipes rather than a tad smaller.
The "design" is really simple. Just find pvc/plastic pipe/fittings that this squeezes into (w/ some tape wrap maybe for assistance). The battery holders for 18650 or whatever can just go on the insides of the pipe just upstream of the HSF. Other than just being simple, it allows for a sliding mechanism for zoom lights if you get some adapters. A 3" FF PVC coupler fits the 75mm aspheric at DX perfectly. Originally I thought about the screwdown pvc pieces would be perfect for setting focus, but they seem much tighter than would be practical, but sanding down pvc friction fit pieces for sliding instead should work ok for occasion show off usage.
Bushytails,
(First off sorry for misspelling your name - I edited those) 
I found a better explanation in this T.I. application note:
http://focus.ti.com/general/docs/litabsmultiplefilelist.tsp?literatureNumber=slva211
Here is the circuit I was trying to describe in my first reply:
Its from the MPS datasheet here: http://www.monolithicpower.com/Page/DownLoad.aspx?ListID=1a4312a6-2709-4b7d-af52-9f1cedbfd415&&ItemID=35
The circuit uses the LED as the voltage reference. Yes, the LED forward Voltage (Vf) of an LED increases with current (non-linearly) but any one LED will have a "known repeatable" Vf at any one current. The Vf will shift lower as the LED heats up, but that should be minor.
So, if you place a voltage divider across the LED, you get a portion of the voltage needed by the feedback pin (0.925V in the case of the MP2307). In this example circuit a 26.1k is already there so adding a 6.04k ohm, as shown, produces 0.625V across the 6.04k resistor when the LED has 3A through it. Add this voltage to the voltage across the 0.1ohm resistor (3A x 0.1ohm = 0.3V), and you get 0.925V at the feedback pin of the IC, and circuit is in regulation.
This technique is not perfect but is simple. It works best when supplying 2/3 (or less) of the required feedback voltage. As always, a lot of testing is needed to make sure it is reliable.
I hope I've made sense here - it's late and the bed is calling...
-Crux
Agenthex,
I get it now! You could rent it out to stores when they have their Grand Opening! LOL
Wow its late...
Crux: Yeah, I figured what you were doing... I just don't like it. :)
That's more of a voltage regulator than a current regulator... even if you disconnected the LEDs entirely, the voltage would barely rise.
I'd be worried about thermal runaway or other problems, and to get accurate output, you'd need to re-tune it for every set of LEDs you try.
By my math, your design puts out 4.92V at no load, but what it puts out with load is highly dependent on the individual Vf of the LED used... a 3.0V led would get 3.6A, a 3.3V led gets 3A, a 3.6V led 2.5A, etc. This is not a very good example of constant current. :)
Using a .33 ohm resistor would make it a true constant-current supply, but would dissipate 3W. Probably the simplest true constant current is to toss a zener and a pull-up resistor in for a reference, or a 7805, depending on what parts you had handy.
--Bushytails
Thanks Crux! I think I'll try this design taking in account your advises. I have some more questions.
1. Is there an issue if those two leds connected to driver in paralell has different Vf? I don't have 3 XM-Ls on one PCB yet, but have three separate ones from different sources to test design.
2. Is there any difference (besides lower current output) if I test design with 4*7135 driver? I have few of them so accidently frying one won't hurt to much. 
3. I can recall someone have mention to transfer excess heat dissipation to resistor which should be last element in the whole circuit (+,1st led, driver (with 2 leds), resistor, -). What do you think about that?
BTW, where did you get that round heatsink with fan? It looks very prommising for cooling highly driven leds.
Ledoman,
- Yes, if you put LEDs with different Vf in parallel the bulk of the current follow "the path of least resistance", the LED with lower Vf. However, the Vf's don't have to be matched exactly for this set-up to work because the Vf increases with current there will be some "spillover" to the other LED. I would measure the voltage across each of your LEDs, taking care to run them at the exact same current, and put the closest two in parallel. I'd do this by running each LED (one at a time) from the 4*7135 driver you have and then measure the voltage across each LED.
- No difference but lower current.
- I'd add heat-sinking to the drivers before I'd add the resistor myself.
- The round heatsink with fan is a ThermalTake from an old PentiumIII processor.